4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2009, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2012, 2017, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
30 * Lustre is a trademark of Sun Microsystems, Inc.
32 * lustre/osd-zfs/osd_internal.h
33 * Shared definitions and declarations for zfs/dmu osd
35 * Author: Alex Zhuravlev <bzzz@whamcloud.com>
36 * Author: Mike Pershin <tappro@whamcloud.com>
37 * Author: Johann Lombardi <johann@whamcloud.com>
40 #ifndef _OSD_INTERNAL_H
41 #define _OSD_INTERNAL_H
43 #include <dt_object.h>
44 #include <md_object.h>
45 #include <lustre_quota.h>
46 #include <lustre_scrub.h>
52 #include <sys/nvpair.h>
53 #include <sys/zfs_znode.h>
56 #include <sys/dmu_objset.h>
57 #include <lustre_scrub.h>
60 * By design including kmem.h overrides the Linux slab interfaces to provide
61 * the Illumos kmem cache interfaces. To override this and gain access to
62 * the Linux interfaces these preprocessor macros must be undefined.
64 #ifdef kmem_cache_destroy
65 #undef kmem_cache_destroy
68 #ifdef kmem_cache_create
69 #undef kmem_cache_create
72 #ifdef kmem_cache_alloc
73 #undef kmem_cache_alloc
76 #ifdef kmem_cache_free
77 #undef kmem_cache_free
80 #define ZFS_VERSION_CODE \
81 OBD_OCD_VERSION(ZFS_MAJOR, ZFS_MINOR, ZFS_PATCH, ZFS_FIX)
83 #define LUSTRE_ROOT_FID_SEQ 0
84 #define DMU_OSD_SVNAME "svname"
85 #define DMU_OSD_OI_NAME_BASE "oi"
87 #define OSD_GFP_IO (GFP_NOFS | __GFP_HIGHMEM)
89 /* Statfs space reservation for grant, fragmentation, and unlink space. */
90 #define OSD_STATFS_RESERVED_SIZE (16ULL << 20) /* reserve 16MB minimum */
91 #define OSD_STATFS_RESERVED_SHIFT (7) /* reserve 0.78% of all space */
93 /* Statfs {minimum, safe estimate, and maximum} dnodes per block */
94 #define OSD_DNODE_MIN_BLKSHIFT (DNODES_PER_BLOCK_SHIFT)
95 #define OSD_DNODE_EST_BLKSHIFT (12) /* est 4KB/dnode */
96 #define OSD_DNODE_EST_COUNT 4096
98 #define OSD_GRANT_FOR_LOCAL_OIDS (2ULL << 20) /* 2MB for last_rcvd, ... */
100 #define OSD_MAX_CACHE_SIZE OBD_OBJECT_EOF
102 #ifndef HAVE_ZFS_REFCOUNT_ADD
103 #define zfs_refcount_add refcount_add
106 extern struct dt_body_operations osd_body_scrub_ops;
107 extern struct dt_body_operations osd_body_ops;
108 extern struct kmem_cache *osd_object_kmem;
111 * Iterator's in-memory data structure for quota file.
113 struct osd_it_quota {
114 struct osd_object *oiq_obj;
115 /* DMU accounting object id */
118 zap_cursor_t *oiq_zc;
119 /** identifier for current quota record */
121 unsigned oiq_reset:1; /* 1 -- no need to advance */
126 OZI_POS_DOT = 1, /* cursor at . */
127 OZI_POS_DOTDOT = 2, /* cursor at .. */
128 OZI_POS_REAL = 3, /* cursor at real entries */
132 * Iterator's in-memory data structure for ZAPs
134 * ZFS does not store . and .. on a disk, instead they are
135 * generated up on request
136 * to follow this format we do the same
139 zap_cursor_t *ozi_zc;
140 struct osd_object *ozi_obj;
141 unsigned ozi_reset:1; /* 1 -- no need to advance */
142 /* ozi_pos - position of the cursor */
143 enum osd_zap_pos ozi_pos;
145 char ozi_name[MAXNAMELEN]; /* file name for dir */
146 __u64 ozi_key; /* binary key for index files */
149 #define DT_IT2DT(it) (&((struct osd_zap_it *)it)->ozi_obj->oo_dt)
152 * regular ZFS direntry
154 struct zpl_direntry {
155 uint64_t zde_dnode:48,
158 } __attribute__((packed));
161 * lustre direntry adds a fid to regular ZFS direntry
163 struct luz_direntry {
164 struct zpl_direntry lzd_reg;
165 struct lu_fid lzd_fid;
166 } __attribute__((packed));
169 /* cached SA attributes */
174 #ifdef ZFS_PROJINHERIT
188 #define OSD_INS_CACHE_SIZE 8
191 struct osd_idmap_cache {
192 struct osd_device *oic_dev;
193 struct lu_fid oic_fid;
194 /** max 2^48 dnodes per dataset, avoid spilling into another word */
195 uint64_t oic_dnode:DN_MAX_OBJECT_SHIFT,
196 oic_remote:1; /* FID isn't local */
199 struct osd_inconsistent_item {
200 /* link into lustre_scrub::os_inconsistent_items,
201 * protected by lustr_scrub::os_lock. */
202 struct list_head oii_list;
204 /* The right FID <=> oid mapping. */
205 struct osd_idmap_cache oii_cache;
207 unsigned int oii_insert:1; /* insert or update mapping. */
210 struct osd_otable_it {
211 struct osd_device *ooi_dev;
212 struct lu_fid ooi_fid;
214 __u64 ooi_prefetched_dnode;
217 /* The following bits can be updated/checked w/o lock protection.
218 * If more bits will be introduced in the future and need lock to
219 * protect, please add comment. */
220 unsigned int ooi_used_outside:1, /* Some user out of OSD
221 * uses the iteration. */
222 ooi_all_cached:1, /* No more entries can be
223 * filled into cache. */
224 ooi_user_ready:1, /* The user out of OSD is
225 * ready to iterate. */
226 ooi_waiting:1; /* it::next is waiting. */
229 extern const struct dt_index_operations osd_otable_ops;
231 /* max.number of regular attributes the callers may ask for */
232 # define OSD_MAX_IN_BULK (sizeof(struct osa_attr)/sizeof(uint64_t))
234 struct osd_thread_info {
235 const struct lu_env *oti_env;
237 struct lu_fid oti_fid;
239 struct ost_id oti_ostid;
245 char oti_key[MAXNAMELEN + 1];
246 __u64 oti_key64[(MAXNAMELEN + 1)/sizeof(__u64)];
247 sa_bulk_attr_t oti_attr_bulk[OSD_MAX_IN_BULK];
249 struct lustre_mdt_attrs oti_mdt_attrs;
250 unsigned int oti_in_trans:1;
252 struct lu_attr oti_la;
253 struct osa_attr oti_osa;
254 zap_attribute_t oti_za;
255 zap_attribute_t oti_za2;
256 dmu_object_info_t oti_doi;
257 struct luz_direntry oti_zde;
259 struct lquota_id_info oti_qi;
260 struct lu_seq_range oti_seq_range;
262 /* dedicated OI cache for insert (which needs inum) */
263 struct osd_idmap_cache *oti_ins_cache;
264 int oti_ins_cache_size;
265 int oti_ins_cache_used;
266 /* inc by osd_trans_create and dec by osd_trans_stop */
267 int oti_ins_cache_depth;
268 struct lu_buf oti_xattr_lbuf;
270 zap_cursor_t oti_zc2;
273 extern struct lu_context_key osd_key;
275 static inline struct osd_thread_info *osd_oti_get(const struct lu_env *env)
277 return lu_context_key_get(&env->le_ctx, &osd_key);
281 struct thandle ot_super;
282 struct list_head ot_dcb_list;
283 struct list_head ot_stop_dcb_list;
284 struct list_head ot_unlinked_list;
285 struct list_head ot_sa_list;
287 struct lquota_trans ot_quota_trans;
291 #define OSD_OI_NAME_SIZE 24
294 * Object Index (OI) instance.
297 char oi_name[OSD_OI_NAME_SIZE];
304 uint64_t *os_compat_dirs;
305 int os_subdir_count; /* subdir count for each seq */
306 u64 os_seq; /* seq number */
307 struct list_head os_seq_list; /* list to seq_list */
310 struct osd_seq_list {
311 rwlock_t osl_seq_list_lock; /* lock for seq_list */
312 struct list_head osl_seq_list; /* list head for seq */
313 struct semaphore osl_seq_init_sem;
316 #define OSD_OST_MAP_SIZE 32
323 struct dt_device od_dt_dev;
324 /* information about underlying file system */
325 struct objset *od_os;
326 uint64_t od_rootid; /* id of root znode */
327 dnode_t *od_unlinked; /* dnode of unlinked zapobj */
328 /* SA attr mapping->id,
329 * name is the same as in ZFS to use defines SA_ZPL_...*/
330 sa_attr_type_t *z_attr_table;
332 struct proc_dir_entry *od_proc_entry;
333 struct lprocfs_stats *od_stats;
335 uint64_t od_remote_parent_dir;
336 uint64_t od_index_backup_id;
337 uint64_t od_max_blksz;
340 struct osd_oi **od_oi_table;
341 unsigned int od_oi_count;
342 struct osd_seq_list od_seq_list;
344 unsigned int od_dev_set_rdonly:1, /**< osd_ro() called */
345 od_prop_rdonly:1, /**< ZFS property readonly */
351 unsigned int od_dnsize;
352 int od_index_backup_stop;
354 enum lustre_index_backup_policy od_index_backup_policy;
361 __s64 od_auto_scrub_interval;
362 struct lu_site od_site;
364 dnode_t *od_groupused_dn;
365 dnode_t *od_userused_dn;
366 #ifdef ZFS_PROJINHERIT
367 dnode_t *od_projectused_dn;
370 /* quota slave instance for inode */
371 struct qsd_instance *od_quota_slave_md;
373 /* quota slave instance for block */
374 struct qsd_instance *od_quota_slave_dt;
376 struct brw_stats od_brw_stats;
377 atomic_t od_r_in_flight;
378 atomic_t od_w_in_flight;
380 /* used to debug zerocopy logic: the fields track all
381 * allocated, loaned and referenced buffers in use.
382 * to be removed once the change is tested well. */
383 atomic_t od_zerocopy_alloc;
384 atomic_t od_zerocopy_loan;
385 atomic_t od_zerocopy_pin;
387 arc_prune_t *arc_prune_cb;
389 /* osd seq instance */
390 struct lu_client_seq *od_cl_seq;
392 struct semaphore od_otable_sem;
393 struct osd_otable_it *od_otable_it;
394 struct lustre_scrub od_scrub;
395 struct list_head od_ios_list;
396 struct list_head od_index_backup_list;
397 struct list_head od_index_restore_list;
399 unsigned long long od_readcache_max_filesize;
402 static inline struct qsd_instance *osd_def_qsd(struct osd_device *osd)
405 return osd->od_quota_slave_dt;
407 return osd->od_quota_slave_md;
410 enum osd_destroy_type {
411 OSD_DESTROY_NONE = 0,
412 OSD_DESTROY_SYNC = 1,
413 OSD_DESTROY_ASYNC = 2,
417 struct dt_object oo_dt;
419 * Inode for file system object represented by this osd_object. This
420 * inode is pinned for the whole duration of lu_object life.
422 * Not modified concurrently (either setup early during object
423 * creation, or assigned by osd_create() under write lock).
426 sa_handle_t *oo_sa_hdl;
427 nvlist_t *oo_sa_xattr;
428 struct list_head oo_sa_linkage;
430 /* used to implement osd_object_*_{lock|unlock} */
431 struct rw_semaphore oo_sem;
433 /* to serialize some updates: destroy vs. others,
434 * xattr_set, object block size change etc */
435 struct rw_semaphore oo_guard;
437 /* protected by oo_guard */
438 struct list_head oo_unlinked_linkage;
440 /* cached attributes */
441 rwlock_t oo_attr_lock;
442 struct lu_attr oo_attr;
444 /* external dnode holding large EAs, protected by oo_guard */
446 enum osd_destroy_type oo_destroy;
448 __u32 oo_destroyed:1,
450 #ifdef ZFS_PROJINHERIT
456 /* the i_flags in LMA */
459 int oo_ea_in_bonus; /* EA bytes we expect */
461 /* record size for index file */
462 unsigned char oo_keysize;
463 unsigned char oo_recsize;
464 unsigned char oo_recusize; /* unit size */
466 uint64_t oo_parent; /* used only at object creation */
468 struct lu_object_header *oo_header;
471 int osd_statfs(const struct lu_env *, struct dt_device *, struct obd_statfs *,
472 struct obd_statfs_info *);
473 extern const struct dt_index_operations osd_acct_index_ops;
474 extern struct lu_device_operations osd_lu_ops;
475 extern struct dt_index_operations osd_dir_ops;
476 int osd_declare_quota(const struct lu_env *env, struct osd_device *osd,
477 qid_t uid, qid_t gid, qid_t projid, long long space,
478 struct osd_thandle *oh,
479 enum osd_quota_local_flags *local_flags,
480 enum osd_qid_declare_flags osd_qid_declare_flags);
481 uint64_t osd_objs_count_estimate(uint64_t refdbytes, uint64_t usedobjs,
482 uint64_t nrblocks, uint64_t est_maxblockshift);
483 int osd_unlinked_object_free(const struct lu_env *env, struct osd_device *osd,
489 static inline int lu_device_is_osd(const struct lu_device *d)
491 return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &osd_lu_ops);
494 static inline struct osd_object *osd_obj(const struct lu_object *o)
496 LASSERT(lu_device_is_osd(o->lo_dev));
497 return container_of(o, struct osd_object, oo_dt.do_lu);
500 static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
502 LASSERT(lu_device_is_osd(&d->dd_lu_dev));
503 return container_of(d, struct osd_device, od_dt_dev);
506 static inline struct osd_device *osd_dev(const struct lu_device *d)
508 LASSERT(lu_device_is_osd(d));
509 return osd_dt_dev(container_of(d, struct dt_device, dd_lu_dev));
512 static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
514 return osd_obj(&d->do_lu);
517 static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
519 return osd_dev(o->oo_dt.do_lu.lo_dev);
522 static inline struct lu_device *osd2lu_dev(struct osd_device *osd)
524 return &osd->od_dt_dev.dd_lu_dev;
527 static inline struct objset * osd_dtobj2objset(struct dt_object *o)
529 return osd_dev(o->do_lu.lo_dev)->od_os;
532 static inline int osd_invariant(const struct osd_object *obj)
538 * Put the osd object once done with it.
540 * \param obj osd object that needs to be put
542 static inline void osd_object_put(const struct lu_env *env,
543 struct osd_object *obj)
545 dt_object_put(env, &obj->oo_dt);
548 static inline int osd_object_invariant(const struct lu_object *l)
550 return osd_invariant(osd_obj(l));
553 static inline struct seq_server_site *osd_seq_site(struct osd_device *osd)
555 return osd->od_dt_dev.dd_lu_dev.ld_site->ld_seq_site;
558 static inline char *osd_name(struct osd_device *osd)
560 return osd->od_svname;
563 static inline void zfs_set_bit(int nr, __u8 *addr)
565 set_bit(nr, (unsigned long *)addr);
568 static inline int zfs_test_bit(int nr, __u8 *addr)
570 return test_bit(nr, (const unsigned long *)addr);
573 static inline int osd_oi_fid2idx(struct osd_device *dev,
574 const struct lu_fid *fid)
576 return fid->f_seq & (dev->od_oi_count - 1);
579 static inline struct osd_oi *osd_fid2oi(struct osd_device *osd,
580 const struct lu_fid *fid)
582 LASSERTF(osd->od_oi_table && osd->od_oi_count >= 1,
583 "%s: "DFID", oi_count %d\n",
584 osd_name(osd), PFID(fid), osd->od_oi_count);
586 return osd->od_oi_table[osd_oi_fid2idx(osd, fid)];
589 #ifdef CONFIG_PROC_FS
591 LPROC_OSD_READ_BYTES = 0,
592 LPROC_OSD_WRITE_BYTES = 1,
593 LPROC_OSD_GET_PAGE = 2,
594 LPROC_OSD_NO_PAGE = 3,
595 LPROC_OSD_CACHE_ACCESS = 4,
596 LPROC_OSD_CACHE_HIT = 5,
597 LPROC_OSD_CACHE_MISS = 6,
598 LPROC_OSD_COPY_IO = 7,
599 LPROC_OSD_ZEROCOPY_IO = 8,
600 LPROC_OSD_TAIL_IO = 9,
604 extern struct kmem_cache *osd_zapit_cachep;
606 extern struct lprocfs_vars lprocfs_osd_obd_vars[];
608 int osd_procfs_init(struct osd_device *osd, const char *name);
609 int osd_procfs_fini(struct osd_device *osd);
612 extern char *osd_obj_tag;
613 int __osd_obj2dnode(objset_t *os, uint64_t oid, dnode_t **dnp);
614 void osd_object_sa_dirty_rele(const struct lu_env *env, struct osd_thandle *oh);
615 void osd_object_sa_dirty_add(struct osd_object *obj, struct osd_thandle *oh);
616 int __osd_obj2dbuf(const struct lu_env *env, objset_t *os,
617 uint64_t oid, dmu_buf_t **dbp);
618 struct lu_object *osd_object_alloc(const struct lu_env *env,
619 const struct lu_object_header *hdr,
620 struct lu_device *d);
621 int osd_object_sa_update(struct osd_object *obj, sa_attr_type_t type,
622 void *buf, uint32_t buflen, struct osd_thandle *oh);
623 int __osd_zap_create(const struct lu_env *env, struct osd_device *osd,
624 dnode_t **zap_dnp, dmu_tx_t *tx, struct lu_attr *la,
625 unsigned dnsize, zap_flags_t flags);
626 int __osd_object_create(const struct lu_env *env, struct osd_device *osd,
627 struct osd_object *obj, const struct lu_fid *fid,
628 dnode_t **dnp, dmu_tx_t *tx, struct lu_attr *la);
629 int __osd_attr_init(const struct lu_env *env, struct osd_device *osd,
630 struct osd_object *obj, sa_handle_t *sa_hdl, dmu_tx_t *tx,
631 struct lu_attr *la, uint64_t parent, nvlist_t *);
632 int osd_find_new_dnode(const struct lu_env *env, dmu_tx_t *tx,
633 uint64_t oid, dnode_t **dnp);
636 int osd_oi_init(const struct lu_env *env, struct osd_device *o);
637 void osd_oi_fini(const struct lu_env *env, struct osd_device *o);
638 int osd_fid_lookup(const struct lu_env *env,
639 struct osd_device *, const struct lu_fid *, uint64_t *);
640 uint64_t osd_get_name_n_idx(const struct lu_env *env, struct osd_device *osd,
641 const struct lu_fid *fid, char *buf, int bufsize,
643 int osd_options_init(void);
644 int osd_ost_seq_exists(const struct lu_env *env, struct osd_device *osd,
646 int osd_idc_find_and_init(const struct lu_env *env, struct osd_device *osd,
647 struct osd_object *obj);
648 struct osd_idmap_cache *osd_idc_find_or_init(const struct lu_env *env,
649 struct osd_device *osd,
650 const struct lu_fid *fid);
651 struct osd_idmap_cache *osd_idc_find(const struct lu_env *env,
652 struct osd_device *osd,
653 const struct lu_fid *fid);
654 int osd_idc_find_and_init_with_oid(const struct lu_env *env,
655 struct osd_device *osd,
656 const struct lu_fid *fid,
658 int fid_is_on_ost(const struct lu_env *env, struct osd_device *osd,
659 const struct lu_fid *fid);
660 int osd_obj_find_or_create(const struct lu_env *env, struct osd_device *o,
661 uint64_t parent, const char *name, uint64_t *child,
662 const struct lu_fid *fid, bool isdir);
664 extern unsigned int osd_oi_count;
667 int osd_get_fid_by_oid(const struct lu_env *env, struct osd_device *osd,
668 uint64_t oid, struct lu_fid *fid);
669 int osd_index_try(const struct lu_env *env, struct dt_object *dt,
670 const struct dt_index_features *feat);
671 int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
672 u64 seq, struct lu_seq_range *range);
673 void osd_zap_cursor_init_serialized(zap_cursor_t *zc, struct objset *os,
674 uint64_t id, uint64_t dirhash);
675 int osd_zap_cursor_init(zap_cursor_t **zc, struct objset *os,
676 uint64_t id, uint64_t dirhash);
677 void osd_zap_cursor_fini(zap_cursor_t *zc);
678 uint64_t osd_zap_cursor_serialize(zap_cursor_t *zc);
679 int osd_remote_fid(const struct lu_env *env, struct osd_device *osd,
680 const struct lu_fid *fid);
681 int osd_add_to_remote_parent(const struct lu_env *env,
682 struct osd_device *osd,
683 struct osd_object *obj,
684 struct osd_thandle *oh);
685 int osd_delete_from_remote_parent(const struct lu_env *env,
686 struct osd_device *osd,
687 struct osd_object *obj,
688 struct osd_thandle *oh, bool destroy);
689 int __osd_xattr_load_by_oid(struct osd_device *osd, uint64_t oid,
693 int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev);
694 void osd_scrub_cleanup(const struct lu_env *env, struct osd_device *dev);
695 int osd_scrub_start(const struct lu_env *env, struct osd_device *dev,
697 void osd_scrub_stop(struct osd_device *dev);
698 int osd_oii_insert(const struct lu_env *env, struct osd_device *dev,
699 const struct lu_fid *fid, uint64_t oid, bool insert);
700 int osd_oii_lookup(struct osd_device *dev, const struct lu_fid *fid,
704 int __osd_sa_xattr_schedule_update(const struct lu_env *env,
705 struct osd_object *obj,
706 struct osd_thandle *oh);
707 int __osd_sa_attr_init(const struct lu_env *env, struct osd_object *obj,
708 struct osd_thandle *oh);
709 int __osd_sa_xattr_update(const struct lu_env *env, struct osd_object *obj,
710 struct osd_thandle *oh);
711 int __osd_xattr_load(struct osd_device *osd, sa_handle_t *hdl,
713 int __osd_xattr_get_large(const struct lu_env *env, struct osd_device *osd,
714 uint64_t xattr, struct lu_buf *buf,
715 const char *name, int *sizep);
716 int osd_xattr_get_internal(const struct lu_env *env, struct osd_object *obj,
717 struct lu_buf *buf, const char *name, int *sizep);
718 int osd_xattr_get_lma(const struct lu_env *env, struct osd_object *obj,
720 int osd_xattr_get(const struct lu_env *env, struct dt_object *dt,
721 struct lu_buf *buf, const char *name);
722 int osd_declare_xattr_set(const struct lu_env *env, struct dt_object *dt,
723 const struct lu_buf *buf, const char *name,
724 int fl, struct thandle *handle);
725 int osd_xattr_set(const struct lu_env *env, struct dt_object *dt,
726 const struct lu_buf *buf, const char *name, int fl,
727 struct thandle *handle);
728 int osd_declare_xattr_del(const struct lu_env *env, struct dt_object *dt,
729 const char *name, struct thandle *handle);
730 int osd_xattr_del(const struct lu_env *env, struct dt_object *dt,
731 const char *name, struct thandle *handle);
732 void osd_declare_xattrs_destroy(const struct lu_env *env,
733 struct osd_object *obj,
734 struct osd_thandle *oh);
735 int osd_xattrs_destroy(const struct lu_env *env,
736 struct osd_object *obj, struct osd_thandle *oh);
737 int osd_xattr_list(const struct lu_env *env, struct dt_object *dt,
738 const struct lu_buf *lb);
739 void __osd_xattr_declare_set(const struct lu_env *env, struct osd_object *obj,
740 int vallen, const char *name, struct osd_thandle *oh);
741 int __osd_sa_xattr_set(const struct lu_env *env, struct osd_object *obj,
742 const struct lu_buf *buf, const char *name, int fl,
743 struct osd_thandle *oh);;
744 int __osd_xattr_set(const struct lu_env *env, struct osd_object *obj,
745 const struct lu_buf *buf, const char *name, int fl,
746 struct osd_thandle *oh);
747 int __osd_sa_xattr_update(const struct lu_env *env, struct osd_object *obj,
748 struct osd_thandle *oh);
750 #define OSD_BASE_EA_IN_BONUS (ZFS_SA_BASE_ATTR_SIZE + \
751 sizeof(__u64) /* VBR VERSION */ + \
752 sizeof(struct lustre_mdt_attrs) /* LMA */)
754 #ifdef HAVE_DMU_OBJECT_ALLOC_DNSIZE
755 int osd_find_dnsize(struct osd_device *osd, int ea_in_bonus);
758 osd_find_dnsize(struct osd_device *osd, int ea_in_bonus)
760 return DN_MAX_BONUSLEN;
764 static inline int osd_object_is_zap(dnode_t *dn)
766 return (dn->dn_type == DMU_OT_DIRECTORY_CONTENTS ||
767 dn->dn_type == DMU_OT_USERGROUP_USED);
770 /* XXX: f_ver is not counted, but may differ too */
771 static inline void osd_fid2str(char *buf, const struct lu_fid *fid, int len)
773 snprintf(buf, len, DFID_NOBRACE, PFID(fid));
777 osd_xattr_set_internal(const struct lu_env *env, struct osd_object *obj,
778 const struct lu_buf *buf, const char *name, int fl,
779 struct osd_thandle *oh)
783 if (unlikely(!dt_object_exists(&obj->oo_dt) || obj->oo_destroyed))
787 if (osd_obj2dev(obj)->od_xattr_in_sa) {
788 rc = __osd_sa_xattr_set(env, obj, buf, name, fl, oh);
790 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
792 rc = __osd_xattr_set(env, obj, buf, name, fl, oh);
798 static inline uint64_t attrs_fs2zfs(const uint32_t flags)
800 return (flags & LUSTRE_APPEND_FL ? ZFS_APPENDONLY : 0) |
801 (flags & LUSTRE_NODUMP_FL ? ZFS_NODUMP : 0) |
802 #ifdef ZFS_PROJINHERIT
803 (flags & LUSTRE_PROJINHERIT_FL ? ZFS_PROJINHERIT : 0) |
805 (flags & LUSTRE_IMMUTABLE_FL ? ZFS_IMMUTABLE : 0);
808 static inline uint32_t attrs_zfs2fs(const uint64_t flags)
810 return (flags & ZFS_APPENDONLY ? LUSTRE_APPEND_FL : 0) |
811 (flags & ZFS_NODUMP ? LUSTRE_NODUMP_FL : 0) |
812 #ifdef ZFS_PROJINHERIT
813 (flags & ZFS_PROJINHERIT ? LUSTRE_PROJINHERIT_FL : 0) |
815 (flags & ZFS_IMMUTABLE ? LUSTRE_IMMUTABLE_FL : 0);
820 #ifndef HAVE_DSL_POOL_CONFIG
821 static inline void dsl_pool_config_enter(dsl_pool_t *dp, void *name)
825 static inline void dsl_pool_config_exit(dsl_pool_t *dp, void *name)
830 #ifdef HAVE_SPA_MAXBLOCKSIZE
831 #define osd_spa_maxblocksize(spa) spa_maxblocksize(spa)
832 #define osd_spa_maxblockshift(spa) fls64(spa_maxblocksize(spa) - 1)
834 #define osd_spa_maxblocksize(spa) SPA_MAXBLOCKSIZE
835 #define osd_spa_maxblockshift(spa) SPA_MAXBLOCKSHIFT
836 #define SPA_OLD_MAXBLOCKSIZE SPA_MAXBLOCKSIZE
839 #ifdef HAVE_SA_SPILL_ALLOC
841 osd_zio_buf_alloc(size_t size)
843 return sa_spill_alloc(KM_SLEEP);
847 osd_zio_buf_free(void *buf, size_t size)
852 #define osd_zio_buf_alloc(size) zio_buf_alloc(size)
853 #define osd_zio_buf_free(buf, size) zio_buf_free(buf, size)
856 #ifdef HAVE_DMU_OBJECT_ALLOC_DNSIZE
857 static inline uint64_t
858 osd_dmu_object_alloc(objset_t *os, dmu_object_type_t objtype, int blocksize,
859 int dnodesize, dmu_tx_t *tx)
862 dnodesize = max_t(int, dmu_objset_dnodesize(os),
865 return dmu_object_alloc_dnsize(os, objtype, blocksize, DMU_OT_SA,
866 DN_BONUS_SIZE(dnodesize), dnodesize, tx);
869 static inline uint64_t
870 osd_zap_create_flags(objset_t *os, int normflags, zap_flags_t flags,
871 dmu_object_type_t ot, int leaf_blockshift,
872 int indirect_blockshift, int dnodesize, dmu_tx_t *tx)
875 dnodesize = max_t(int, dmu_objset_dnodesize(os),
878 return zap_create_flags_dnsize(os, normflags, flags, ot,
879 leaf_blockshift, indirect_blockshift,
880 DMU_OT_SA, DN_BONUS_SIZE(dnodesize),
885 osd_obj_bonuslen(struct osd_object *obj)
887 int bonuslen = DN_BONUS_SIZE(DNODE_MIN_SIZE);
889 if (obj->oo_dn != NULL && obj->oo_dn->dn_num_slots != 0) {
890 bonuslen = DN_SLOTS_TO_BONUSLEN(obj->oo_dn->dn_num_slots);
892 objset_t *os = osd_dtobj2objset(&obj->oo_dt);
896 dnodesize = dmu_objset_dnodesize(os);
898 bonuslen = DN_BONUS_SIZE(dnodesize);
905 static inline uint64_t
906 osd_dmu_object_alloc(objset_t *os, dmu_object_type_t objtype, int blocksize,
907 int dnodesize, dmu_tx_t *tx)
909 return dmu_object_alloc(os, objtype, blocksize, DMU_OT_SA,
910 DN_MAX_BONUSLEN, tx);
913 static inline uint64_t
914 osd_zap_create_flags(objset_t *os, int normflags, zap_flags_t flags,
915 dmu_object_type_t ot, int leaf_blockshift,
916 int indirect_blockshift, int dnodesize, dmu_tx_t *tx)
918 return zap_create_flags(os, normflags, flags, ot, leaf_blockshift,
919 indirect_blockshift, DMU_OT_SA,
920 DN_MAX_BONUSLEN, tx);
924 osd_obj_bonuslen(struct osd_object *obj)
926 return DN_MAX_BONUSLEN;
928 #endif /* HAVE_DMU_OBJECT_ALLOC_DNSIZE */
930 #ifdef HAVE_DMU_PREFETCH_6ARG
931 #define osd_dmu_prefetch(os, obj, lvl, off, len, pri) \
932 dmu_prefetch((os), (obj), (lvl), (off), (len), (pri))
934 #define osd_dmu_prefetch(os, obj, lvl, off, len, pri) \
935 dmu_prefetch((os), (obj), (lvl), (off))
938 static inline int osd_sa_handle_get(struct osd_object *obj)
940 struct osd_device *osd = osd_obj2dev(obj);
941 dnode_t *dn = obj->oo_dn;
947 dbuf_read(dn->dn_bonus, NULL, DB_RF_MUST_SUCCEED | DB_RF_NOPREFETCH);
948 rc = -sa_handle_get_from_db(osd->od_os, &dn->dn_bonus->db, obj,
949 SA_HDL_PRIVATE, &obj->oo_sa_hdl);
952 zfs_refcount_add(&dn->dn_bonus->db_holds, osd_obj_tag);
956 static inline void osd_dnode_rele(dnode_t *dn)
960 LASSERT(dn->dn_bonus);
963 dmu_buf_rele(&db->db, osd_obj_tag);
966 #ifdef HAVE_DMU_USEROBJ_ACCOUNTING
968 #define OSD_DMU_USEROBJ_PREFIX DMU_OBJACCT_PREFIX
969 #define OSD_DMU_USEROBJ_PREFIX_LEN DMU_OBJACCT_PREFIX_LEN
971 static inline bool osd_dmu_userobj_accounting_available(struct osd_device *osd)
973 return dmu_objset_userobjspace_present(osd->od_os);
977 #define OSD_DMU_USEROBJ_PREFIX "obj-"
978 #define OSD_DMU_USEROBJ_PREFIX_LEN 4
980 static inline bool osd_dmu_userobj_accounting_available(struct osd_device *osd)
984 #endif /* #ifdef HAVE_DMU_USEROBJ_ACCOUNTING */
986 static inline int osd_zap_add(struct osd_device *osd, uint64_t zap,
987 dnode_t *dn, const char *key,
988 int int_size, int int_num,
989 const void *val, dmu_tx_t *tx)
993 #ifdef HAVE_ZAP_ADD_BY_DNODE
995 return -zap_add_by_dnode(dn, key, int_size, int_num, val, tx);
997 return -zap_add(osd->od_os, zap, key, int_size, int_num, val, tx);
1000 static inline int osd_zap_remove(struct osd_device *osd, uint64_t zap,
1001 dnode_t *dn, const char *key,
1006 #ifdef HAVE_ZAP_ADD_BY_DNODE
1008 return -zap_remove_by_dnode(dn, key, tx);
1010 return -zap_remove(osd->od_os, zap, key, tx);
1014 static inline int osd_zap_lookup(struct osd_device *osd, uint64_t zap,
1015 dnode_t *dn, const char *key,
1016 int int_size, int int_num, void *v)
1020 #ifdef HAVE_ZAP_ADD_BY_DNODE
1022 return -zap_lookup_by_dnode(dn, key, int_size, int_num, v);
1024 return -zap_lookup(osd->od_os, zap, key, int_size, int_num, v);
1027 static inline void osd_tx_hold_zap(dmu_tx_t *tx, uint64_t zap,
1028 dnode_t *dn, int add, const char *name)
1030 #ifdef HAVE_DMU_TX_HOLD_ZAP_BY_DNODE
1032 dmu_tx_hold_zap_by_dnode(tx, dn, add, name);
1036 dmu_tx_hold_zap(tx, zap, add, name);
1039 static inline void osd_tx_hold_write(dmu_tx_t *tx, uint64_t oid,
1040 dnode_t *dn, uint64_t off, int len)
1042 #ifdef HAVE_DMU_TX_HOLD_ZAP_BY_DNODE
1044 dmu_tx_hold_write_by_dnode(tx, dn, off, len);
1048 dmu_tx_hold_write(tx, oid, off, len);
1051 static inline void osd_dmu_write(struct osd_device *osd, dnode_t *dn,
1052 uint64_t offset, uint64_t size,
1053 const char *buf, dmu_tx_t *tx)
1056 #ifdef HAVE_DMU_WRITE_BY_DNODE
1057 dmu_write_by_dnode(dn, offset, size, buf, tx);
1059 dmu_write(osd->od_os, dn->dn_object, offset, size, buf, tx);
1063 static inline int osd_dmu_read(struct osd_device *osd, dnode_t *dn,
1064 uint64_t offset, uint64_t size,
1065 char *buf, int flags)
1068 #ifdef HAVE_DMU_READ_BY_DNODE
1069 return -dmu_read_by_dnode(dn, offset, size, buf, flags);
1071 return -dmu_read(osd->od_os, dn->dn_object, offset, size, buf, flags);
1075 #ifdef HAVE_DMU_OBJSET_OWN_6ARG
1076 #define osd_dmu_objset_own(name, type, ronly, decrypt, tag, os) \
1077 dmu_objset_own((name), (type), (ronly), (decrypt), (tag), (os))
1079 #define osd_dmu_objset_own(name, type, ronly, decrypt, tag, os) \
1080 dmu_objset_own((name), (type), (ronly), (tag), (os))
1083 #ifdef HAVE_DMU_OBJSET_DISOWN_3ARG
1084 #define osd_dmu_objset_disown(os, decrypt, tag) \
1085 dmu_objset_disown((os), (decrypt), (tag))
1087 #define osd_dmu_objset_disown(os, decrypt, tag) \
1088 dmu_objset_disown((os), (tag))
1092 osd_index_register(struct osd_device *osd, const struct lu_fid *fid,
1093 __u32 keysize, __u32 recsize)
1095 return lustre_index_register(&osd->od_dt_dev, osd_name(osd),
1096 &osd->od_index_backup_list, &osd->od_lock,
1097 &osd->od_index_backup_stop,
1098 fid, keysize, recsize);
1102 osd_index_backup(const struct lu_env *env, struct osd_device *osd, bool backup)
1104 struct lu_fid *fid = &osd_oti_get(env)->oti_fid;
1107 lu_local_obj_fid(fid, INDEX_BACKUP_OID);
1108 rc = osd_idc_find_and_init_with_oid(env, osd, fid,
1109 osd->od_index_backup_id);
1113 lustre_index_backup(env, &osd->od_dt_dev, osd_name(osd),
1114 &osd->od_index_backup_list, &osd->od_lock,
1115 &osd->od_index_backup_stop, backup);
1118 #ifndef HAVE_DMU_TX_MARK_NETFREE
1119 #define dmu_tx_mark_netfree(tx)
1122 #ifndef HAVE_ZFS_INODE_TIMESPEC
1123 #define inode_timespec_t timestruc_t
1126 #endif /* _OSD_INTERNAL_H */